Capturing the attention of readers? Stylistic and psychological ...

Capturing the attention of readers?

Stylistic and psychological perspectives on the use and

effect of text fragmentation in narratives*

CATHERINE EMMOTT, ANTHONY J. SANFORD, AND LORNA I. MORROW

Abstract

This article brings together researchers from Stylistics and Psychology to studywhether text fragmentation, which appears often to be used by writers as a fore-grounding device (Mukařovský 1964), is able to capture the attention of readersof narratives. We examine two types of text fragmentation: sentence fragmentsand mini-paragraphs. Firstly, we study the stylistic functions of fragmentation, in-cluding its cumulative use at plot crucial moments and its use for local rhetoricalpurposes. We then turn to psychological research on depth of processing (e.g.Sanford and Sturt 2002) and introduce a new method of testing, the text changedetection method (Sturt et al. 2004). We report an experiment using this methodto examine whether text fragments and very short sentences can increase theamount of detail that readers notice in a text, and then discuss the results in rela-tion to potential applications. The work provides both a case study of the empir-ical analysis of foregrounding devices, and also, more generally, a case study ofinter-disciplinary research across the Humanities and Social Sciences.

0. Introduction: Bridging Stylistics and Psychology – The inter-disciplinary study of text fragmentation and attention capture

Literary scholars have recognized the crucial role that a reader’s attention canplay in the interpretation of a text and the way in which a writer can use linguis-tic features to attract attention. Rabinowitz (1987: 53), for example, has sug-gested, in a chapter strikingly entitled “Trumpets, please!: Rules of notice” thata text has:

[…] a hierarchical organization of details: we do not attend to everything equally […]the stressed features in a text serve as a basic structure on which to build an interpre-tation […] we read with the prior understanding that we are more expected to accountfor a detail that is stressed by a rule of notice than for a detail that is not.

JLS 35 (2006), 1–30 0341–7638/06/035– 1DOI 10.1515/JLS.2006.001 © Walter de Gruyter

2 Catherine Emmott et al.

The idea that parts of a text can be “foregrounded” is one of the core assump-tions within Stylistics (Mukařovský 1964; see also van Peer 1986 and van Peerand Hakemulder 2006). Wales (2001: 157) defines foregrounding as both (i) “the“throwing into relief” of the linguistic sign against the background of the normsof ordinary language” and (ii) “within the literary text itself linguistic featurescan themselves be foregrounded, or “highlighted”, “made prominent”, for spe-cial effects, against the (subordinated) background of the rest of the text”. Mostliterary scholars and stylisticians either draw on their intuitions to judge whatbreaks the norms, or perform statistical analyses of the frequency of particularlinguistic features in specific texts and/or in corpora of texts (e.g. Leech andShort 1981; Stubbs 2005). These text-based approaches take for granted that lin-guistic norm-breaking has an effect on readers, but do not provide any evidencethat readers’ attention is attracted and do not assess the amount of impact that“foregrounded” features have on readers. By contrast, our approach in this arti-cle is to supplement traditional stylistic analysis with an examination of these sty-listic intuitions by testing the response of readers to unusual features in texts.Empirical research of foregrounding devices has been pioneered by Humanitiesresearchers in the area termed the “Empirical Study of Literature” (e.g. van Peer1986; Steen 1994), in relation to topics such as the interpretation of metaphors inpoetry. The research described in this article brings together Humanities and So-cial Science researchers from Stylistics and Psychology to examine a specific is-sue, whether changes to the usual sentence and paragraph formatting of narra-tive texts can change the readers’ alertness to details in texts. In our stylisticanalysis, we focus on cases where texts suddenly become fragmented, with unu-sually frequent sentence or paragraph breaks. Our stylistic intuitions are thatfragmentation can capture the attention of readers; the aim of our empiricalwork is to show whether this is really the case. Our research tests fundamentalstylistic assumptions about foregrounding, and is also a case study in inter-disci-plinary research across the Humanities and Social Sciences.

The stylistic discussion in the first part of this article examines the followingtwo types of text fragmentation:

(i) Sentence fragments – These are words, phrases or clauses that cannot beclassed as syntactically full sentences, but are nevertheless punctuated graphi-cally as sentences.1 The two final noun phrases in the following example aretherefore sentence fragments:2

(1) It was only dinner. One dinner. One guest.(Monica Ali [2004]. Brick Lane, p. 19)

The interest in sentence fragments dates back to the early days of modern sty-listics (e.g. Crystal and Davy 1969: 49–51; Leech and Short 1981: 214–219)where their study was part of a more general interest in the structure of sentenc-

Capturing the attention of readers? 3

es, particularly in relation to how the simplicity or complexity of a sentencecould serve stylistic purposes. Such sentence fragments break prescriptivists’expectations about the well-formedness of sentences in written text. Neverthe-less, they are used widely for rhetorical and stylistic purposes across a widerange of genres (e.g. advertising language, newspaper writing, popular sciencetexts, literary texts and popular fiction) and have been of particular interest toresearchers in speech and thought representation in fiction (e.g. Fludernik1993: 235). They are also acknowledged in grammar books (e.g. Quirk et al.1985: 1446; Huddleston and Pullum 2002: 944–945, 1735) and even in recentwriting style guides (e.g. Allen 2002: 57 and King 2004: 13–14). In spoken dis-course, the common use of fragments has been used to challenge the assump-tion that there is any underlying notion of sentences as we speak (e.g. Miller andWeinert 1998: Chapter 2). One significant feature of fragments is that they aregenerally brief. We investigate this brevity in the empirical work in Section 3.2by testing them alongside very short sentences.

(ii) Mini-paragraphs – In this case, blocks of text are fragmented by unusuallyfrequent paragraph breaks. In principle, a mini-paragraph is any paragraph thatis exceptionally short. Clearly, statistical analysis would be necessary to ascer-tain the precise cut-off point for a paragraph being considered exceptionallyshort in a corpus and/or specific texts. Rather than perform this type of statisti-cal analysis here, our analysis focuses almost entirely on examples which are in-disputably short. These are “single item” mini-paragraphs, where the paragraphcontains a single sentence or a single fragment.3 Example 2 shows a single sen-tence mini-paragraph in an autobiographical narrative, and Example 3 shows asentence fragment in a mini-paragraph in the preface of a novel:

(2) Early in the winter of 1975, I’d planned a road trip to some desert climbs inJoshua Tree National Park in southern California. Most experienced climb-ers headed to Yosemite in the summer, but I wasn’t quite ready for its hugegranite walls. For me at that stage, the smaller granite climbs in the desertwould be a good first step away from the sandstone of Eldorado SpringsCanyon. I couldn’t wait.

Two weeks before the trip, my father died.I’d known from my mother that he had throat cancer. But since leaving

Greece with the rest of the family, I’d seen him only once in six years, forall of five minutes, when I was fourteen years old.(David Breashears [2000]. High Exposure: An Enduring Passion for Ever-est and Unforgiving Places, pp. 62–63)

(3) The Scheme for Full Employment was the envy of the world: the greatestundertaking ever conceived by men and women. It solved at a stroke theproblem that had beset humankind for generations. Participants had only


to put the wheels in motion, and they could look forward to a bright, sunlitupland where idleness and uncertainty would be banished forever. Plannedto the finest detail by people of vision, The Scheme was watertight, andcould not possibly go wrong.

Except in this country.In this country we managed to destroy it. We destroyed the last thing that

could save us from obscurity and ruin. And we did it with our own handstoo, not at the behest of some errant leader whom we could obey and thenblame later.(Magnus Mills [2003]. The Scheme for Full Employment, unnumberedpreface)

Although the role of paragraphs has been studied by academics (e.g. Longacre1968) and is mentioned in writing style guides (e.g. King 2004: 15), there seemsto have been relatively little sustained study of unusual paragraph breaks, ex-cept in bibliography studies of lay-out, authors’ revisions and editing practices(e.g. Adamson 1995; Bray et al. 2000; Sotirova 2004).

The stylistic functions of fragmented text will be examined in Sections 1 and 2below. Sentence fragments and mini-paragraphs often appear to draw attentionto the information they contain by segmenting off specific pieces of informationfrom the rest of the text. To study this, we will be drawing on research in Psychol-ogy on how the attention of a reader can be “captured” (a term derived fromwork on visual attention [e.g. Simons 2000a; Franconeri et al. 2005]). Our re-search is carried out against a background of psychological work on depth ofprocessing (as summarized in Sanford and Sturt 2002), which will be discussedin more detail in Section 3.1. This body of work provides a substantial amount ofevidence that readers of texts are normally quite inattentive to the details ofwhat they read. As will be illustrated in Section 3.1, readers often do not fully en-gage with specific words in a text and do not form a full semantic representationof what they read. This means that readers are generally quite poor at spottingerrors and contradictions in texts. By using devices that capture the attention ofreaders, it is possible to increase the extent to which they process the text and im-prove the accuracy of their semantic representations. This aspect of the work hasobvious implications for literacy research and has already been used to investi-gate the differences between skilled and less-skilled readers (e.g. Hannon andDaneman 2004). So far, linguistic devices such as clefting have been studied bypsychologists for their ability to raise attention levels, but no psychological workhas been done previously on the fragmentation devices discussed here. The factthat fragmentation has not previously been studied by psychologists is perhapsnot surprising since mainstream cognitive psychologists normally base their hy-potheses on artificial examples of their own creation and these examples arehighly unlikely to include the types of stylistic feature discussed in this article.4


The shifts in attention levels studied by psychologists in depth of processingresearch can be quite subtle, although nevertheless crucial to readers’ under-standing of texts. By contrast, in stylistic studies, the interest in attention gener-ally tends to be in cases where the text features are so obviously unusual that theycan be expected to make a difference to key aspects of readers’ literary interpre-tation of texts. Our stylistic analysis began from this perspective, with an interestin how extremely heavy cumulative foregrounding devices, such as repeated useof fragmentation, were used at crucial points in the plot of a story with the appar-ent intention of forcing the reader to become aware of highly significant infor-mation or objects in a story or encouraging the reader to make a key plot infer-ence. Then, as we studied fragmentation in more detail, we looked more broadlyat the range of uses of these devices in narratives, some of which may have morelocal stylistic and rhetorical effects of emphasis. In addition to these uses, for thepurposes of this inter-disciplinary article, we are broadening our view of levels ofattention further to cases where a sentence fragment or mini-paragraph may beof interest simply because its simplicity makes a difference to the amount a read-er notices rather than because a writer seems to be overtly using it for the pur-poses of emphasis. Although this latter area may be of less direct interest to sty-listicians, it may, as mentioned above, be of considerable relevance to literacyresearch, in areas such as Educational Linguistics, English as a Foreign Lan-guage, and Languages for Specific Purposes.

Overall, our project provides an unusual bridge between the Humanities re-search in Stylistics and Science research in Psychology, enabling us to bring newideas and methods into each discipline. The research on depth of processing pro-vides a new angle for many stylisticians since any stylistician who works by simplyanalysing the text can only speculate about the amount that a reader notices, par-ticularly for the less stylistically marked examples. We are also able to draw onnew testing methods, in this case a methodology developed recently by Sanfordand his colleagues (e.g. Sturt et al. 2004), the text change detection method (seeSection 3.2 for further details). The empirical work described in this article is onlya first step to understanding how stylistic features capture the attention of thereader, but we nevertheless believe that it initiates a programme of study whichwill provide empirical evidence for fundamental stylistic assumptions. At thesame time, studies of stylistic phenomena serve to enhance psychological theoriesof the process of reading. We are using our observations about narrative texts toinform the design of our empirical materials, including stylistic features in thesematerials that would never normally be studied by Cognitive Psychologists.

The main body of this article is divided into the following sections. In Sec-tions 1 and 2, we focus on the stylistic analysis, providing a more detailed stylis-tic survey of text fragmentation. In Section 1, we examine some of the graphicalfeatures which initially captured our own attention as researchers, looking athow heavy foregrounding devices are used at key plot moments in narratives.


Section 2 broadens our observations about text fragmentation, giving a summa-ry of some of the key stylistic and rhetorical uses of sentence fragments andmini-paragraphs across narrative texts generally. In this stylistic discussion, ourspeculations about the stylistic functions of fragmentation and our use of termssuch as “foregrounding” and “highlighting” do, of course, prejudge the issue ofwhether these graphical features have any effect on readers. Our stylistic anal-ysis is typical of the subjective approach in the Humanities, where scholars offera plausible interpretation of how a reader might respond to a text. By contrast,the approach in Psychology is to check even the “taken for granted” assump-tions about reading, providing detailed models of the psychological processesinvolved. In Section 3, we switch to a psychological analysis of text fragments,in an attempt to find evidence of whether fragmentation does in fact capture theattention of readers. Section 3.1 provides the theoretical background in relationto depth of processing research. Section 3.2 introduces the text change detec-tion method and gives the empirical report of our first experiment on fragmen-tation. This experiment also tests the reading of very short sentences to see theeffect of brevity generally on attention capture. In Section 4, we make some sty-listic observations about the empirical results and discuss possible applications.

1. Stylistic analysis (I): Cumulative use of fragmentation at key plot moments in narratives

In our discussion of narrative examples in Sections 1 and 2, we draw on literarytexts, popular fiction, (auto)biographies, and embedded narrative within popu-lar science writing. The texts from which our examples have been selected aregenerally reasonably conventional in their sentence and paragraph structure,but on occasions they suddenly become fragmented for specific stylistic purpos-es. We have deliberately chosen in this article not to examine the work of writ-ers who have reputations for being highly innovative in their use of lay-out andtypefaces (e.g. Irvine Welsh and Janice Galloway) since our interest here is inhow fragmentation supplements the normal resources for writing, rather thanin deliberately experimental usage of graphical features.5

Our stylistic analysis in this section shows the cases where text fragmentationseems most clearly to be being used by writers with the intention of capturing theattention of readers. In each of the following examples, cumulative fragmenta-tion is used to present key plot information. In Example 4, there is a combinationof a sentence fragment and a mini-paragraph. In Example 5, there is a string offragments with an unusual paragraph break between two (almost) repeatedphrases. In Example 6, there is a string of sentence fragments around a short sen-tence, all in mini-paragraphs. In Example 7, there are several sentence fragmentsused to summarize a particular episode, and the line beginning “Later” seems, inour judgement, sufficiently brief to be classed as a mini-paragraph. The fore-


grounding seems so blatant in these cases, that we refer to the cumulative use ofthe fragmentation and other devices as potential “attention grabbers”.

(4) Everything was automatic now – down the stairs to the cellar, the lightswitch, the deep freeze, the hand inside the cabinet taking hold of the firstobject it met. She lifted it out, and looked at it. It was wrapped in paper, soshe took off the paper and looked at it again.

A leg of lamb.All right then, they would have lamb for supper.

(Roald Dahl [1979]. “Lamb to the slaughter”, p. 25)

(5) Rebus could see a bed, a figure on the bed, [hospital] staff milling aroundthe various machines […] There was clothing folded on a chair, a bag be-neath it. As the doctor pulled out the bag, Rebus saw something. A flatwhite cardboard box.

A white cardboard pizza box. Clothes: black denims, black bra, red satinshirt. A black duffel-coat. […](Ian Rankin [2000]. The Hanging Garden, pp. 18–19)

(6) And then, on a list dated 1992, [Kathryn] saw the name she hadn’t even re-alized she’d been looking for, the unusual name that rose right up from thepaper and traveled through her bones with a charge.

Muire Boland.Flight attendant.Kathryn spoke the name aloud.Muire Boland.

She was pretty sure it was a woman’s name. She wondered if it was French andwhether she was pronouncing it correctly. Kathryn reached down in front ofher and opened the large drawer of Jack’s desk. The junk-mail envelope withthe name penciled in a corner wasn’t there, but she could see it just as clearly asshe could see the typed name on the list she held in her hands. Muire 3.30, thehastily scrawled note had read. On an envelope, a solicitation from Bay Bank.(Anita Shreve [1999]. The Pilot’s Wife, p. 182, Shreve’s italics)

(7) ‘In our garden,’ says Linda. ‘A couple of years after she was married, Ithink. She [Kath] dropped in, quite out of the blue – off to see friends inCornwall. But I gathered she’d just had that nasty little upset, poor dear, soshe was rather under par.’ Linda gives Elaine a furtive look – regret, andcomplicity. ‘Such a shame –’

[One page of text omitted]Later, she hears Linda again. That nasty little upset. Under par.But later is on hold until after six, when the garden empties, the customers

and the cars depart, Jim and his nephew pile into the red truck, Pam goes offto the pub with an admirer from the village. Alone, Elaine checks the ticket


sales and is quietly pleased with visitor levels, locks up the shop, goes into thesilent house. Only then does that other aspect of the afternoon come bustlingin. Cousin Linda hangs around all evening, saying that again. And again.(Penelope Lively [2004]. The Photograph, p. 161 and p. 162)

In each of these examples the text fragmentation is used for a piece of informa-tion which ultimately has key plot significance, but which might not appear im-portant at the point of reading unless highlighted in some way (see also Emmott[2003] for a discussion of how these types of feature can add to a plot). In theDahl story (Example 4), the leg of lamb subsequently becomes the murderweapon in the story. This provides the crux of the “twist in the tale” since, ironi-cally, the policemen investigating the crime later eat the leg of lamb whilst delib-erating over the whereabouts of the unidentified weapon. In Example 5, from adetective story by Ian Rankin, the mention of the pizza box provides a key cluewhich identifies a previously unidentified character. In the Shreve story (Exam-ple 6), readers of the previous part of the novel will know that there are reasonsfor Kathryn to suspect that her dead husband, a pilot, may have had an affairwith one of his colleagues. As she scans a list of staff who worked with him andnotices the name “Muire Boland”, she remembers unexplained personal jottingsthat she has previously found in his belongings, including the reminder “Muire3.30”. In the Lively story (Example 7), the mention of “that nasty little upset”and Kath having being “under par” could just refer to some minor ailment, butthe “furtive look” of the speaker suggests that there is more to it than this. Later,we find that this is a euphemistic way of referring to a miscarriage, an event thatis significant since it provides an explanation for Kath’s suicide.

The foregrounding has different purposes in these narratives. In the Dahl sto-ry (Example 4), the reader is not aware until shortly after this extract that theleg of lamb will be used as a murder weapon. This extract (together with the ti-tle of the story) highlights an object that is about to become of crucial impor-tance. In Examples 5 and 6, the foregrounding provides sufficient informationfor crucial inferences to be made by readers of the full stories up to these points.In Example 5, the Rankin detective novel, the mention of the pizza box tells thereader that Rebus is looking not at an unidentified accident victim, but at hisdaughter, who was last seen by him carrying a pizza box. In the Shreve novel,the reader of the story has already been given sufficient information for it to bepossible to start drawing inferences about the role that Muire Boland had in thedead husband’s life as the stretch of text in Example 6 is read. In the Lively nov-el, the mention of “that nasty little upset” in Example 7 provides a sense of mys-tery throughout the following pages until the nature of the “upset” is explained.

In these types of plot-crucial examples, the cumulative fragmentation doesnot occur in isolation, but is accompanied by linguistic features that are charac-teristic of these heavily foregrounded text extracts, as follows:


(i) Italics

Example 6 shows the use of italics in addition to fragmentation. (See also Em-mott et al. (2003: 22) for a very similar example from detective fiction where thecombination of fragmentation and italicisation is used to highlight a crucial cluein the story.) Italics are well known as a graphical device that can be used forthe purposes of emphasis. Previous research in Sanford’s laboratory (McAteer1989, 1992) has shown that italics can have an effect on readers and, more re-cently, we have used the text change detection method described in Section 3.2to show that italics can alter the extent to which readers notice and rememberdetails in a text (Alison J. S. Sanford et al. in press).

(ii) Under-specification

Under-specification relates to cases where there is a lack of identifying informa-tion about a referent. The notion of under-specification is an idea currently be-ing explored in Cognitive Science (in areas such as Psychology and Computing)which is in line with classic work within Literary Theory, such as Ingarden’s(1973) “spots of indeterminacy”, Iser’s (1978) “gaps” or “blanks”, and Barthes’(1974) “enigmas”. Under-specification may sometimes occur simply because acharacter or object is of such minor importance that extra details are irrelevantor because the focalizing character is unaware of or does not remember the fullinformation (see Emmott 2006 for examples). On other occasions, under-spec-ification may combine with explicit statements of significance or textual fore-grounding devices to create interest in the missing information6. Examples 4–7all use under-specification to a greater or lesser extent. Example 7 is the mostobvious since the main fragmented detail (“That nasty little upset.”) is itself un-der-specified. In Examples 4, 5 and 6, there is a delay before the relevant objectsare identified, with the items being initially under-specified as “the first object”,“something”, and “the name … the unusual name”, respectively. It may be thatunder-specification has the potential to focus the reader’s attention on the itembeing referred to, while the reader awaits further specification.

(iii) Repeated references to the same item without any significant narrative development

In all these examples there is repetition, with the narrative lingering on specificdetails with little or no narrative development. In Example 4, the under-speci-fied item (“the first object … it … it … It … it”) is eventually named in the frag-ment “A leg of lamb.” and then the final line refers again to “lamb”. In Example5, the near-repetition in the sentence fragments “A flat white cardboard box.”and “A white cardboard pizza box.” serves both to highlight this crucial itemand also to add a new piece of information (that it is a pizza box) to make the


clue directly link in with a previous description of the same item many pagesearlier. In Example 6, there is repetition of the reference to the discovered“name” (“the name”, “the unusual name”, “the name”, “a woman’s name” and“the typed name”), in addition to repetition of “Muire Boland”. In Example 7,the fragments “That nasty little upset.” and “Under par.” are echoing the pre-vious conversational utterance. This last example includes an odd slowing andre-iterating of the narrative presentation, since we are told that the re-play hap-pens “Later”, but then that “later is on hold”. When “later” occurs, the repeti-tion of “again” iconically represents the repeated echoing of the phrase in thecharacter’s mind.

In Examples 4–7, the foregrounding techniques are clearly closely linked withthe representation of the focalizers’ thoughts. If we assume that the ultimate pur-pose of these passages is to highlight plot-crucial information for the reader, thefact that the linguistic devices can also be interpreted as showing the characters’thought processes can add to the plausibility of the passages. In Example 4, thehighlighting of the leg of lamb might seem very heavy-handed if it were not forthe fact that the presentation also shows the perceptual process of the characteras she gradually recognizes this item in her freezer. Focalization may also makeclearer that the graphical features are indicating significant information. Theremay be explicit statements of the considerable importance of the relevant object/information to a character and therefore to the reader, as in Example 6 in the de-scription of the shock that Kathryn experiences as she reads the name that “trav-el[s] through her bones with a charge”. More subtly, the preoccupation of the fo-calizer with the discovered item (as represented by the lingering repetitiondescribed above in Examples 5, 6 and 7, in particular) may also serve to show thatthe item is highly significant to the character and therefore to the reader too.

The cumulative use of fragmentation, often coupled with the above addition-al linguistic devices, appear to serve as “attention grabbers”, which we might as-sume to prompt a very high level of attention capture. In the above examples,these devices seem to be emphasizing information which otherwise could easilybe overlooked since there is no explicit statement in any of these cases of whythe information is of any particular relevance to the focalizer (e.g. in Example6, we have to infer that the Muire Boland was the focalizer’s husband’s lover).We will see in Section 2 that text fragments can also occur individually, but thenthey may be emphasizing local details (Section 2.1) or may provide additionalhighlighting for more explicitly-stated globally-relevant information (Section2.2). One hypothesis stemming from our stylistic analysis is that it may be nec-essary to use fragmentation devices and other highlighting devices cumulativelyin order to increase attention in cases where the reader needs to infer the globalsignificance of information. In such cases, heavy cumulative use of graphicalfeatures may hint to the reader that the text has greater importance than mayappear at first sight.


2. Stylistic analysis (II): Text fragmentation from a broader perspective

Section 1 shows the plot-central use of cumulative sentence fragmentation andmini-paragraphs which first attracted our stylistic interest, but our study hassince broadened to look at each of these features in their range of uses in a sur-vey of 30 narratives (novels, biographies, short stories and narrative embeddedin popular science texts). For reasons of space, we present here only a selectionof examples from this study, discussing only some of the more common usesthat we found (see Emmott et al. 2006 for many more examples of sentencefragmentation and for a wider range of categories).

There are several reasons for extending the stylistic analysis in this way. First-ly, this broader perspective shows that fragmentation has a far wider range ofpurposes than the plot-central foregrounding discussed in Section 1, hence in-dicating that the study of fragmentation is of more general interest than just atnarrative climaxes. Secondly, although we will again see examples of cumula-tive fragmentation, a significant number of our examples will include individualinstances of fragmentation, so we can explore whether such individual instancesalso appear to have a foregrounding role. Thirdly, a broader range of examplesprovides an indication of the different ways that fragmentation can be integrat-ed into a text (e.g. showing how fragments can echo or expand on the previoustext). In Section 2.1, we focus on key uses of sentence fragmentation and in Sec-tion 2.2, we examine some of the main functions of mini-paragraphs.

2.1. Sentence fragmentation

The following list gives some of the common uses of sentence fragmentation. Inmany of these cases, both the brevity of the fragment and the fact that it is splitoff by punctuation could be argued to be a strategy for placing local emphasison the information in the fragment. If so, then we might expect these forms tocapture the attention of readers, as will be investigated by the empirical workin Section 3.2.

(i) Repetition and paraphrase

(8) His face worked relentlessly, writhing with anger and dread. Mostly anger.(Paul Broks [2004]. Into the Silent Land, p. 18)

(9) But standing there, she suddenly had the sense that Robert Hart was al-ways observant. Always watching.(Anita Shreve [1999]. The Pilot’s Wife, p. 167)

Sentence fragments commonly echo words that have occurred in the precedingsentence. The fragment may repeat a word or phrase, often with additional in-tensification (Example 8), or provide a paraphrase (Example 9). The content it-


self may provide emphasis (i.e. by re-iteration and intensification), but the frag-mented form also serves to make the information stand on its own, arguablyhighlighting this material further.

(ii) Additional detail and greater specificity

(10) Then she remembered, her stomach lurching, the twice-yearly trainingsessions in London. Two weeks each.(Anita Shreve [1999]. The Pilot’s Wife, p. 226)

(11) Catching sight of me she smiled, and at that moment the boy turned andglanced back at her. He had floppy, blond hair and sad, drooping eyes.Connolly.(Zoë Heller [2004]. Notes on a Scandal, p. 158)

The fragment can add extra detail which might otherwise have appeared in theprevious sentence (Example 10). Sometimes, initial underspecification of thereferent can be followed by a more precise label, such as a name, often simulat-ing the thought processes of a character as he or she gradually discovers infor-mation (Example 11). Here, we might hypothesize that by placing the additionalinformation alone in the sentence fragment, it might highlight the informationmore than if it had been included within the previous text. This is not just a mat-ter of the length of the fragment, but syntactic structure can also be relevant. If“two weeks”, in Example 10, was embedded inside the earlier noun-phrase (e.g.“the twice-yearly two-week training sessions”), then we might expect the em-bedded information to have less prominence by contrast. We will return to thispoint about embedding in the discussion of depth of processing in Section 3.1.

(iii) Fractured structures

(12) And I, with the eventual support of my parents, went back to Syracuse.Alone.(Alice Sebold [2002]. Lucky, p. 88)

(13) Money was just a score, that was all, the mark of your ingenuity. And luck.(Louise Dean [2004]. Becoming Strangers, p. 185)

(14) Only an inattentive daughter could fail to see. A disrespectful daughter.Who fully deserved the lashing, verbal or otherwise, that followed suchdereliction of duty.(Monica Ali [2004]. Brick Lane, p. 205)

In Examples 12 and 13, words which appear to belong to the end of the previoussentence occur as fragments, as if the previous sentence has been fractured intotwo. In Example 14, the relative clause in the second fragment could have oc-curred as a post-modifier of the preceding noun phrase, but the whole phrase has


been fractured into two. These “fractured structures” are recognized in the lead-ing grammar books. Quirk et al. (1985: 1446) suggest that the made-up example“I saw Miriam. And Walter.” gives the impression that Walter is either being men-tioned as an afterthought or that the sight of him is being given special or dramaticsignificance, depending on the prosody attached to the fragment in silent reading.Huddleston and Pullum (2002: 1735) create the example “He had broken thevase. Deliberately.”, suggesting that the punctuation might give extra importanceto the fragment in the same way that prosody could set it apart in speech.

(iv) Evaluations and conclusions

(15) There were two Marthas, and there was nothing to bridge them. Failure.Complete failure.(Doris Lessing [1966]. A Proper Marriage, p. 164)

(16) The great, sun-dried dead church, the dead portales, the hopeless coveredmarket-place, where, the first time she went, she saw a dead dog lying be-tween the meat stalls and the vegetable array, stretched out as if for ever,nobody troubling to throw it away. Deadness within deadness.(D. H. Lawrence [1996]. “The woman who rode away”, p. 39)

Fragments can also embody evaluations, as in Example 15, or can provide con-clusions, as in Example 16. These particular examples are notable for theirrhythmic qualities in the original texts. Example 15 is part of a longer stretch oftext which alternates short fragments and a short sentence with longer sentenc-es. In Example 16, the fragment rounds off the paragraph, providing a short fi-nal summary after the lengthy preceding sentence.

(v) Negatives and repairs

(17) Most Autograph Men have given up the hope of ever getting one. NotAlex.(Zadie Smith [2003]. The Autograph Man, p. 66)

(18) Connor mentioned the wife, and the three children and the dog, fairlysoon after he and Julie met. Well, not met. Slept together. It was almostthe same thing.(Margaret Atwood [1992]. “The bog man”, p. 88)

The fragment can provide a contrast with the previous sentence, for example toshow the difference between a general situation and a specific case (Example 17).Fragments can also provide repairs, sometimes following a negation (Example18). Arguably, placing the negative or repair in a fragment alone may put more em-phasis on the negation or the repaired information than if a longer structure hadbeen used (e.g. with the negation fronted in the fragment in Example 17 ratherthan longer statements such as “Alex had not given up hope.” or “Alex had not.”).


(vi) Orientation and topic management

(19) Tijuana. The border. A three-hour wait at Customs. The landscape chang-es again […](Kenneth A. Brown [1994]. Cycles of Rock and Water: Upheaval at the Pa-cific Edge, p. 34)

(20) And Clare, always Clare. Clare in the morning, sleepy and crumple-faced.Clare with her arms plunging into the papermaking vat, pulling up themold and shaking it so, and so, to meld the fibres. Clare reading, with herhair hanging over the back of the chair, massaging balm into her crackedred hands before bed. Clare’s low voice is in my ear often.(Audrey Niffenegger [2005]. The Time Traveler’s Wife, p. 4)

So far, the fragments discussed have echoed or interacted with the previous text,but it is also possible for them to preface what is to follow. Fragmentation can beused to set the scene at the start of an episode, as in Example 19. Listing the tem-poral-spatial information in fragments can be argued to put more emphasis onthose orientational coordinates than if they were “diluted” by placing these de-tails in longer narrative sentences (e.g. “We arrived at Tijuana.”). In Example 20,the initial fragment provides a “header” for all the statements that follow.

In all of the above examples, a case can be made for fragmentation provid-ing emphasis. In types (i)–(v), the fragment arguably puts more focus on theinformation than if the information had just been included in the previous sen-tence. In type (vi), the fragment provides a means of concentrating on the keyorientation and topic words, arguably giving them more prominence than ifthey had been in a longer sentence. Writers may, nevertheless, use fragmentsfor a variety of purposes, not all of these necessarily being for reasons of em-phasis. Fragmentation is frequently used in direct speech to show elliptical re-sponses and may then simply be mimicking the fragmented nature of realspeech.7 Another very common use is for the fragments to embody a list ofitems or a list of impressions.8 Also, in iconic representations, the purpose ofthe sentence fragments may be to echo physical effects such as speed of move-ment, as shown in Example 21.

(21) The city shattered. Everything was in pieces. She knew it straight away,glimpsed it from the painful-white insides of the ambulance. Frantic neonsigns. Headlights chasing the dark. An office block, cracked with light.These shards of the broken city.(Monica Ali [2004]. Brick Lane, p. 117)

The fact that writers may not always be using emphasis as a deliberate stylisticdevice does not, however, undermine the psychological investigation of frag-mentation as a potential attention-capturing device. Even in cases where pro-viding emphasis does not seem to be the main intention of the writer, psycho-


logical methods can still be used to test whether the fragmentation has thesecondary effect of making information in the fragments more noticeable toreaders and/or more memorable. This might be useful for applications such astesting the readability of texts in literacy research.

2.2. Mini-paragraphs

Mini-paragraphs have specific features which may explain why writers oftenappear to be using them with the intention of highlighting particular lines in atext. Firstly, their brevity may be a significant factor, particularly since the lay-out may have a physical effect on readers. Where there are very frequent par-agraph breaks, it might be the case that this serves to slow down the readingprocess, but that would need to be empirically tested. Likewise, it is possiblethat the extra white space itself serves to capture the attention of the reader,but again this needs to be tested (see Section 3.2). Secondly, the mini-para-graph (particularly the “single item” mini-paragraph) may often express justone idea, rather than bringing together several ideas as we might normally ex-pect in a paragraph, with the possible effect of emphasizing the information inthe mini-paragraph.

Previous examples show two reasons why writers may use mini-paragraphs:plot relevance and rhetorical shifts. In Section 1, we saw in Examples 4–7 thatcumulative fragmentation could be used to present plot-crucial details, appar-ently emphasizing the text so that plot inferences could be made and/or key in-formation highlighted for later use. Even one instance of text fragmentation,such as a mini-paragraph containing a full sentence, can be used to highlightplot information. In Example 22, the mini-paragraph highlights informationthat is of obvious importance to any reader of the full novel, the discovery thatthe main protagonist’s husband had had a son by another woman.

(22) The baby in Muire Boland’s arms, despite the different sex, despite theslightly darker hair colour, was precisely the baby that Mattie had been atthat age – five months old, Kathryn guessed. The realization created dis-sonance, a screeching in her ears, as though this woman she had never metwere holding Kathryn’s child.

Jack had had a son.The dark-haired woman turned and left the hallway for a sitting room,

leaving Kathryn to follow. The child in the hallway, a beautiful girl withenlarged pupils and a cupid mouth, picked up a handful of constructionblocks, pressed them to her chest, and, eyeing Kathryn the entire time,edged along the wall and entered the sitting room, moving closer to hermother’s legs. The girl looked like her mother, whereas the boy, the son,resembled the father.(Anita Shreve [1999]. The Pilot’s Wife, pp. 212–213)


Examples 2 and 3 (in the Introduction) showed rhetorical shifts. In Example 2,the reader has to infer the discontinuity, assuming that the death of the narra-tor’s father would significantly affect his plans to carry out his intended trip. InExample 3, there is a specific indicator of discontinuity (“Except”). The para-graphing is unusual in this case because the first paragraph ends on a positivenote and we would normally expect the paragraph break to signal a switch to anew topic. Instead, the concessive phrase suddenly provides a reversal. Argua-bly, the paragraph break makes the change in direction more surprising than ifthe phrase “except in this country” had been part of the preceding paragraph.

Mini-paragraphs are also used for text which has a range of other functions,often reflecting a shift in hierarchical level where there is a comment on the pre-vious text or a prediction of how the narrative will continue. This applies par-ticularly to cases where a mini-paragraph appears as a coda, a final comment orstatement at the end of an episode and/or a chapter (Labov 1972). In SueTownsend’s (1989) Rebuilding Coventry, for example, one chapter describes thenarrator desperately and unsuccessfully begging for food from a shopkeeper,then seeing the shopkeeper just throw the food away. The chapter ends with themini-paragraph coda “He could have given it to me after all.” (p. 62). The mini-paragraph can also set specific events in a broader context. In Michael J. Fox’s(2003) autobiography, Lucky Man, he ends a chapter with the coda, “The ride,it turned out, was only beginning.” (p. 112), predicting the direction that the re-mainder of the narrative will take. Even within an episode, a mini-paragraphcan provide a switch from specific events to a more general comment. In LanceArmstrong’s (2001) autobiography, It’s not about the Bike: My Journey back toLife, he describes how he repeatedly dismissed the pre-diagnosis signs of can-cer, providing a mini-paragraph comment from his subsequent more knowingstandpoint, that “I had an excuse for everything.” (p. 8). The mini-paragraphcan also be used for heavily evaluative comments which set specific events inthe context of the narrator’s lifetime experiences, as in the following example:

(23) Next morning, at 5:30 A.M., Jangbu violently shook my tent and said,“David, David, get up, you must come to Rob’s camp.” In the blue diningtent that doubled as Adventure Consultants’ communications tent, Edand I found a distraught Finnish climber named Veikka Gustafsson. LikeEd, he had a close friendship with Rob and had climbed other Himalayanpeaks with him. Veikka had been up all night monitoring the radio for anyfurther news. He told us that an hour earlier, Rob had broken the terriblesilence to ask, “Won’t somebody come and help me?”

It was one of the most chilling moments of my life.Rob was a powerful man. He was eminently capable, organized, and

widely experienced. He functioned superbly at high altitude. If anyonebelonged on this mountain, it was Rob, always in command of himself.


And yet he’d lost control.(David Breashears [2000]. High Exposure: An Enduring Passion for Ev-erest and Unforgiving Places, p. 263)

In all of the above examples in this section, the mini-paragraphs might be ar-gued to emphasize crucial information or a key comment on the preceding text.Mini-paragraphs may also be used simply for iconic purposes. In the followingexample, Philip K. Dick uses a mini-paragraph to reflect the slowness and diffi-culty of movement as a character struggles to climb a staircase, as each actiondrains him of his life. The mini-paragraph in Example 24 does not emphasizekey information, but instead echoes the nature of the action described.

(24) His last labored actions governed by a tropism. An orientation urging himtoward death, decay and nonbeing. A dismal alchemy controlled him; cul-minating in the grave.

He ascended another step.I’m going to make it, he realized. The force goading me is feasting on

my body; that’s why Wendy and Al and Edie – and undoubtedly Zafskyby now – deteriorated physically as they died, leaving only a discardedhusklike weightless shell, containing nothing, no essence, no juices, nosubstantial density. The force thrust itself against the weight of many grav-ities, and this is the cost, this using up of the waning body.(Philip K. Dick [2001]. Ubik, p. 184)

As discussed in the previous section, even in cases where fragmentation does nothave the rhetorical function of providing emphasis, psychologists can test wheth-er the use of fragmented structures, such as the mini-paragraph here, increasesthe attention that a reader pays to the information in the structure (in terms ofmaking the details more noticeable). This may have implications for judging thereadability of a text and/or understanding the creation of iconic effects.

3. Psychological analysis: An empirical investigation of text fragmentation

In the previous sections, we have presented a stylistic analysis of some of thekey functions of short sentence fragments and mini-paragraphs used in narra-tive texts. In many (but not all) these cases, it can be argued that a writer ap-pears to be using fragmentation to provide emphasis, thereby aiming to attractthe attention of readers. In this section we look at how the assumptions under-lying this stylistic analysis can be tested empirically, using techniques from ex-perimental psychology. We present our first foray into the problem, describingan initial experiment which tests the following hypotheses. This experiment ex-amines whether sentence fragments can increase the amount of attention thata reader pays to a text, in terms of the amount of detail that the reader notices.


In addition to sentence fragments, our data included some very short sentences(3–5 words), since we felt that brevity generally might be a significant factor inexplaining attention capture. Our experiment also examines whether cumula-tive use of these devices, in the form of a short fragment/sentence plus a mini-paragraph, has a greater effect on the reader’s attention, again in terms of thedetail noticed, than a short fragment/sentence alone.

3.1. Psychological research on attention: Depth of processing

The possibility that structures such as sentence fragments, very short sentences,and mini-paragraphs may raise the level of attention complements previous workon attention by psychologists. In cognitive psychology, it is generally thought thatwhen something is attended to, it requires more substantial processing. For exam-ple, if we pay attention to a particular word, then its meaning will play a greaterrole in our ongoing interpretations. Within Psychology, this area of study is called“depth of processing”.

Depth of processing researchers have found that readers are often remarkablyunobservant about what they read, although to rather different degrees depend-ing on the syntactic and semantic complexity of the reading materials. Readerssee the words in a sentence, but they do not necessarily draw all the inferencesthat could be drawn. The full semantic information does not seem to be being uti-lized in such cases and this is referred to as “shallow depth of processing”.

Shallow depth of processing has been studied particularly in relation to se-mantic anomalies in sentences such as the following9 (e.g. Erickson and Matte-son 1981; Barton and Sanford 1993; Sanford 2002; Sanford and Sturt 2002):

(i) After an aircrash, where should the survivors be buried?(ii) Can a man marry his widow’s sister?(iii) How many animals of each sort did Moses put in the Ark?

Readers often try to answer these questions, failing to spot that they are non-sensical or erroneous because (i) survivors are not usually buried (ii) a man whohas a widow must be dead and so cannot marry, and (iii) it was Noah rather thanMoses who put the animals in the Ark.10 For question (i), for example, readersoften answer “bury them where their relatives want” and for question (iii),readers frequently answer “two” rather than “none”. It seems that readers areaware of the plausibility of a particular word being used in a particular situation(e.g. the word “survivors” is likely to occur in a piece of text describing an air-crash), since they notice more changes when a word is used that is quite inap-propriate to the context (van Oostendorp and de Mul 1990) or when the contextis changed to make a particular word less plausible (Barton and Sanford 1993).Nevertheless, readers are “shallow processing” in the sense that they are oftennot attending to whether every semantic feature of a word fits the context ex-actly (i.e. survivors are alive and so are unlikely to be buried). This shows that


many people do not retrieve from memory the full details of the relevant word(survivors, Moses, etc.), i.e. what the word denotes.

From the point of view of our current research, a key point is that certaintypes of syntactic structure can make semantic anomalies more observable.Hence cleft structures, as in (ii) below, can focus the reader’s attention on theerroneous information and lead to a higher error detection rate than for (i) be-low (e.g. Bredart and Modolo 1988):

(i) Moses put two of each sort of animal in the Ark. True or false?(ii) It was Moses who put two of each sort of animal in the Ark. True or false?

The typical (though not inevitable) answer to question (i) is “true”, with peopleoften not noticing the anomaly, as discussed above. For question (ii), however,very few people say that the sentence is true. The sentence in (ii) is effectivelyan answer to the question “Who was it that put two of each kind of animal onthe Ark?”, that is, Moses is asserted and not presupposed. In contrast, (i) hasmuch broader focus, being an answer to a question like “What happened?”, andMoses is presupposed, not asserted. Focus is a way of manipulating what is at-tended to most in a sentence and clearly influences the degree to which a criticalword is processed. (For related work exploring this idea in greater detail, seeSanford and Sturt 2002 and Sturt et al. 2004.)

Depth of processing can also be decreased by embedding information in sub-ordinate clauses. This has the effect of making anomalies less detectable. In thefollowing sentences containing erroneous information about the liver beingfound only in humans, the anomaly was more difficult to detect in the subordi-nate structure in (i) than in the main clause position in (ii) (e.g. Baker and Wag-ner 1987):

(i) Is the following true or false? The liver, which is an organ found only in hu-mans, is often damaged by heavy drinking.

(ii) Is the following true or false? The liver, which is an organ often damagedby heavy drinking, is found only in humans.

Our current work contributes to this body of research on depth of processing,testing whether sentence fragments, very short sentences, and mini-paragraphscan increase the attentiveness of readers. This is a simple hypothesis, but ifsupported by the experiments, it has the potential to provide an insight intothe way that attention works and add to the above body of work on how read-ers respond to the complexity of sentence structure. Just as cleft structures canmake readers more attentive to certain parts of a text, the graphical featurespresented in this article may be able to do the same. Fragmentation, by its verynature, allows items to appear in isolation that would otherwise have occurredin longer grammatical structures. Sometimes this may be a case of separatingan item from other clause elements that the item would normally have oc-


curred with. A fragment may also be an element that might otherwise have ap-peared as a modifier in a phrase, or a subordinate clause that would otherwisehave occurred embedded in a phrase or another clause. In Example 14, for in-stance, the relative clause “Who fully deserved the lashing, verbal or other-wise, that followed such dereliction of duty.” is extracted from what would oth-erwise have been an embedded position in a noun phrase. The depth ofprocessing research described above has shown that embedding can makereaders process in a shallower fashion, probably because embedded informa-tion is typically presupposed and so is uncontentious. By contrast, it may bethat processing is deeper when a structure that would otherwise be embeddedoccurs as a separate fragment.

3.2. Report of the experimental investigation: The text change detection method in relation to text fragmentation

Our experiment used the new technique of text change detection, developed bySanford and his colleagues (e.g. Sturt et al. 2004). The idea for this method camefrom earlier use of the change detection method in vision research, where sub-jects were tested to see how easily they could spot differences between almostidentical pictures in order to assess how much attention they were paying to de-tails in the pictures (see reviews of this work in Simons and Levin 1997 and Si-mons 2000b). Often changes were not spotted and the phenomenon was re-ferred to as “change blindness”. Instead of pictures, the text change detectionmethod uses pairs of texts which are either identical or identical apart from achange of one word, in order to test whether readers notice that a change hasoccurred. In the experiment described below, by using versions of the textwhich have different graphical formats, we can test the extent to which thegraphical format makes a difference to noticeability of a change.

Using this method, participants read pieces of text through once, then re-read either the same text (about 50% of the time) or the same text with just oneword changed (the remainder of the time). The very simple idea is that if read-ers were attending to a particular word in detail, they would recognize that achange had occurred more often than if they were not attending.

(i) Method: We created 36 short passages, each of which appeared in threeversions. These passages included the types of stylistic features found in ouranalysis of examples from literary texts, popular fiction, (auto)biographiesand narrative embedded in popular science writing. Each of the three ver-sions for one of these passages is shown below. These different graphical for-mats comprise the different conditions of the study. In each case, the bracket-ed material ((→ white)) indicates the word that would be changed in a second


presentation in order to test whether readers noticed changes to a greater orlesser extent in the different conditions.

In the first condition, the critical region (“a brown (→ white) envelope”), waspresented assimilated into an earlier sentence:

Condition 1: Assimilated into an earlier sentencePeter had tried hard not to go over details of the interview in his mind. Hetried not to rephrase his answers, tormenting himself with things he shouldhave said. He tried hard especially not to listen out for the postman everymorning, but at last there it was, a brown (→ white) envelope.

He examined it for a while, before summoning the courage to look insideat his fate.

In the second condition, the critical region was presented as a separate sentencefragment or very short sentence, but did not have its own mini-paragraph:

Condition 2: Short fragment/sentence, no mini-paragraphPeter had tried hard not to go over details of the interview in his mind. Hetried not to rephrase his answers, tormenting himself with things he shouldhave said. He tried hard especially not to listen out for the postman everymorning, but at last there it was. A brown (→ white) envelope.

He examined it for a while, before summoning the courage to look insideat his fate.

In the third condition, the critical region was presented as a separate sentencefragment or very short sentence in its own mini-paragraph:

Condition 3: Short fragment/sentence, mini-paragraphPeter had tried hard not to go over details of the interview in his mind. Hetried not to rephrase his answers, tormenting himself with things he shouldhave said. He tried hard especially not to listen out for the postman everymorning, but at last there it was.

A brown (→ white) envelope.He examined it for a while, before summoning the courage to look in-

side at his fate.

Three computer files were constructed, each containing all 36 experimentalpassages. Within each file, one third of the passages appeared in each of thethree conditions described above. For a given passage, the text appeared in oneof the conditions in the first file, in a second condition in the second file, and ina third condition in the third file. In this way, each passage appeared only oncein each file, but taking all three files into account, each passage appeared in allconditions. By showing only one file to a given participant, we could thus guar-antee that the participant only saw a given passage once, but saw 12 passages ineach condition.


To each file we added 112 “filler” passages, comprising three or four sentenc-es. These were designed to dilute the experimental passages so that participantswould not come to expect their particular structures, which might then influ-ence their reading. In the filler passages, changes to a word occurred in 68 of the112 passages. These changes were distributed over the sites where they oc-curred in the passages, ranging from a word early in the first sentence to a wordlate in the last sentence, again to avoid participants spotting where changesmight occur. To facilitate this, 20 of the fillers had very obvious changes in wide-ly distributed places so that they were easily detected. Finally, a simple compre-hension question was added after the second version of each of the passages tofacilitate reading-for-meaning.

Twenty-four University of Glasgow undergraduates took part. Eight partic-ipants saw the first file, eight the second, and eight the third. The passageswere presented on a screen by computer. Participants first fixated their eyeson a fixation cross (positioned where the start of the passage would appear),and when they were ready pressed a button on the response box to call up thefirst passage. Reading was entirely self-paced. They were instructed to readthe passage naturally, and when they had finished, to press the button again.After a 500 milli-second delay,11 the passage was presented for the secondtime, either with a word change or not. Participants read the passage again,pressing the button when they had finished. At this point, the screen wentblank. The participants then pressed a button on the response box to indicatewhether or not they had detected a change. They were also instructed that ifthey had noticed a change to report verbally to the experimenter what theythought this change was. To start the next trial, they pressed the button againwhen they were ready.

Results: The critical measure is the proportion of changes detected under thethree conditions, and this is shown as a function of condition in Figure 1.

Statistical analysis, using analysis of variance, showed that there were reliabledifferences in detection rate over these conditions, by participants [F1 (2, 46) =4.392, MSe = 361.829, p<.05] and materials [F2 (2, 70) = 6.320, MSe = 366.384,p<.01].12 Newman-Keuls tests were used to examine this main effect in moredetail, revealing that changes were detected significantly less often in the con-dition where the critical item was made part of the earlier sentence (Condition1 = 25.28%) than they were when the item was made to stand alone, either inthe same paragraph (Condition 2 = 41.53%) [q1=4.19, p<.05; q2=5.01, p<.01], orin its own mini-paragraph (Condition 3 = 34.17%) (although this reached sig-nificance in the analysis by items only) [q1=2.29, p>.05; q2=2.90, p<.05]. Therewas no reliable difference between the two stand-alone conditions, i.e. Condi-tions 2 and 3.13


Psychological discussion: These results show that when sentence fragments orvery short sentences are introduced towards the end of short texts (Conditions2 and 3), they enhance detection of changes to words, compared with the sameitem appearing as part of the preceding sentence (Condition 1). We can con-clude that more careful, deeper, processing is occurring for short stand-aloneitems, just as we have generally been assuming in the stylistic analysis. It is in-teresting that we did not find an even greater enhancement of processing in the“mini-paragraph” position, considering the use of the mini-paragraph as a sty-listic device. One possibility is that simply putting the text into its own sentencefragment or very short sentence may be enough to maximize attention, bringingit to a peak. Alternatively, it might be suggested that the result is an artefact ofthe method, because the change detection task is not sensitive enough to detectany additional capture of attention – however, we think this explanation is un-likely because the detection rates were not at 100%, even in the conditionswhere detection was boosted.15 If there genuinely is no increase in attentioncapture in the mini-paragraph case, what role could the mini-paragraph serve?One possibility which will be the focus of future investigation is that a mini-par-agraph may increase reading time duration. This may give rise to further effects,such as increasing the long-term memorability of the item (Waugh 1963).

4. Stylistic discussion

The above experimental results support our earlier stylistic analysis in relationto our assumption that sentence fragments were being used by writers to draw

20

25

30

35

40

45

50

Condition 1:assimilated into an

earlier sentence

Condition 2:short fragment/sentence,

no mini-paragraph

Condition 3:short fragment/sentence,

mini-paragraph

presentation condition

% c

orre

ct c

hang

e de

tect

ion

Figure 1. % correct detection of changes to targets in the critical regions under the threeconditions, including standard error bars14


the attention of readers towards specific parts of texts. This may be due to thefact that these fragments are usually brief,16 an assumption supported by the factthat the very short sentences had the same effect as the fragments. It might alsobe because a fragment effectively “unembeds” information that would other-wise be embedded in or part of a longer structure (as discussed in Section 3.1), afinding that would complement existing psychological work on the contrastingphenomenon of readers’ low levels of attention for embedded information.

These findings have important implications. Depth of processing is a key con-cept because it challenges the traditional psychological idea that readers forma fully-specified representation at each level of text processing. A fully-speci-fied analysis would necessitate understanding the full meaning of every word ina sentence, then combining these words into an interpretation of a sentence. Bycontrast, the psychological evidence for shallow processing shows that a readercan proceed through a text with anomalous words going unnoticed, althoughthis is less likely to be the case when “attention capturing” features, such as clef-ting, and sentence fragmentation are used. From a practical point of view, thistype of research could be used to show how information can be buried in com-plex structures (for example, in the analysis of how politicians can bury detailsin political texts, as studied within Critical Discourse Analysis) or can be usedto show how writers bring details to the forefront of attention in texts such as inadvertisements17 and newspaper articles, since these texts commonly make useof fragmentation. Whether or not writers are deliberately using particular syn-tactic or graphical formats, the research also has considerable potential to beused to show the readability of a text and/or to show how attentively readers ofdifferent abilities read formats of different types. Barton and Sanford’s (1993)work on depth of processing has already been tested by applied psychologists.Hannon and Daneman (2004) found differences between individual readers.Both skilled and less-skilled readers formed plausible representations of partic-ular scenarios, but skilled readers were much better at integrating the semanticdetail embodied in specific words into their representation of the scenario. Interms of the materials discussed in Section 3.1, this would mean that we wouldexpect both skilled and less-skilled readers to be able to recognize that the word“survivors” fits an aircrash scenario, but less-skilled readers would be particu-larly poor at noticing that the essential semantic property of aliveness did notmatch the verb “buried”.18

The results for sentence fragments and very short sentences are, therefore, asignificant step forward, but the results for the placement of these items in amini-paragraph are not as predicted. We expected the mini-paragraph to raiseattentiveness further and this did not happen. The results of the change detec-tion study show no difference between the two conditions where the criticalitem stands alone (i.e. appearing in a mini-paragraph or not). Superficially, theimplication would be that it makes no difference to readers whether, say, Roald


Dahl (in Example 4) had placed the phrase “A leg of lamb.” on a separate lineor not. However, it seems strange that a graphic device like putting an item inits own paragraph does not enhance detection. If there really is no difference,many writers are using an ineffective device and a very fundamental assump-tion of stylistic analysis is open to question. However, an alternative possibilityis to look more carefully at the nature of attention and at what the text changedetection method actually tests. The text change detection method has beenvery successful for showing how much attention readers are giving to specificwords in a text, but the reader’s attention can also be directed towards the sig-nificance of a whole statement. As we saw in Section 2.2, mini-paragraphs innatural text are often used to show a change in rhetorical direction, a change inthe pattern of events, or a shift to a different hierarchical level. In Example 2,for instance, the sentence “Two weeks before the trip, my father died.” providesa sudden change in the expected pattern of events. If the attention of the readeris focused on the impact that this piece of information is likely to have on thenarrator’s plans, then it is possible that it may make little difference to the read-er whether the death occurred two weeks or, say, three weeks before the trip(although we would expect a difference if it had read “Two years before thetrip…”). Indeed, if the reader’s attention is on the overall significance of themini-paragraph, then it may even be that small changes of this type are less no-ticeable. This is, of course, just speculation, but it provides the basis for furtherwork on this topic.

5. Conclusion

Our work in this article provides a case study in inter-disciplinary research. Wehave shown how we have added fresh perspectives to both our stylistic and psy-chological analyses by bringing together ideas and methodologies from thesevery different disciplines. The empirical research has been prompted by our sty-listic analysis and the results add to a significant body of psychological work ondepth of processing. As discussed above, depth of processing research is of the-oretical importance in understanding how we read and it has significant practi-cal applications in areas such as literacy studies and stylistic/rhetorical analysis.We have shown here how the reader’s attention can be captured by placing in-formation in short sentence fragments or very short sentences. Our work is alsoa major step forward in terms of making psychologists more aware of the im-portance of style in the design of empirical materials. However, the case studyalso makes clear that this type of inter-disciplinary collaboration does not al-ways readily produce empirical results which easily mesh with our stylistic intu-itions. The mini-paragraphs require additional testing by other methods beforewe can provide any real answer to how this particular graphical feature affectsreaders. This latter result may, nevertheless, be revealing in the long term about


the nature of attention. The finding for mini-paragraphs suggests that we mayneed to develop a more complex model that accounts for both the reader’s abil-ity to notice local details and the reader’s awareness of how the presentation ofa text can affect the more global rhetorical shape of that text. Although this in-itial experiment has rather mixed results, we therefore see it as an importantfirst step in developing a programme of inter-disciplinary research which wehope will provide major insights into the way that readers attend to specific fea-tures in texts in the process of reading.

University of Glasgow

Notes

* This research was funded by an Arts and Humanities Research Board “InnovationAward” (B/IA/AN8799/APN13740) and a British Academy grant (SG-36219) forour project “Literature, Narrative and Cognitive Science: Interdisciplinary Perspec-tives on the Nature of Reading” (The LINCS Project), 2002-3, and an Arts andHumanities Research Council research grant (B/RG/AN8799/APN17330) for ourproject, “Stylistics, Text Analysis, and Cognitive Science: Interdisciplinary Perspec-tives on the Nature of Reading” (The STACS Project), 2005-8. The authors are grate-ful to the AHRB/C and the British Academy for their support.

1. By “sentence fragment” we mean simply that the fragment is punctuated as a sentencewould be, not that there is any underlying sentence. Some analysts use the term “minorsentence” (e.g. Crystal and Davy 1969: 49–51) for fragments, but we prefer to see thefragments as below the level of the sentence, rather than as types of sentence.

2. Sentence fragments and mini-paragraphs, where applicable to our discussion, areindicated by our added bold throughout the article.

3. The 3-item “mini-paragraph” (“Later …”) in Example 7 provides an exception. Inthat case, we judge the paragraph to be unusually short, but statistical analysis wouldbe necessary to confirm how far it deviates from the average paragraph length in therelevant text and/or in a corpus.

4. See Emmott (1997: Chapter 3) for further discussion of this point.5. See Short (2004) for a stylistic analysis of unusual graphical features in experimental

writing.6. See Toolan (2004) for similar comments about how withholding information about

an event or an individual can create narrative interest. Toolan uses continuation teststo explore these reader expectations.

7. See Emmott et al. (2006) for examples.8. See Emmott et al. (2006) for examples.9. These examples are commonly referred to as “Moses-type illusions” by psychologists,

named after the best-known example of this type (see (iii) below in the main text).10. In this depth of processing research, the experimenters check the general knowledge

and linguistic knowledge of the experimental subjects. Hence, they ensure that sub-jects are eliminated from the statistical analysis if they have insufficient biblical


knowledge to know who put the animals in the Ark, or if they do not know the mean-ing of key words such as “survivor”.

11. This is a standard procedure in most text change detection tests, in order to stop par-ticipants spotting a change purely on the basis of surface visual characteristics.Although experimenters try to make changed words the same length as the originalwords, this is not always possible and, if a second text is presented immediately, par-ticipants may notice a slight flicker as the length of line changes. Also a “flickereffect” can occur simply because everything has stayed the same apart from thechanged word. The 500 milli-second delay is intended to prevent this occurring.

12. Statistics were carried out by examining reliability across subjects (how well theeffect generalises to all participants), indexed by F1 and q1, and by materials (howwell the effect generalises to all materials), indexed by F2 and q2.

13. Since the error bars of the final two columns overlap in Figure 1, the two columns areinterpreted as not being reliably different.

14. The error bars are a statistical calculation of the margin of error and hence show therange over which results might be expected to occur if the experiment were replicated.

15. Failures to detect changes were in fact quite high. The absolute values obtained inchange detection experiments depend upon a variety of factors, including the size ofthe change in meaning from initial word to new word (Sturt et al. 2004). In thepresent experiment, semantic changes were kept quite small.

16. Although fragments are usually brief, they are not necessarily so (e.g. the relativeclause in Example 14 is not particularly short).

17. Sanford and colleagues have previously conducted some empirical work in relation tothe style of writing used in product information. Sanford et al. (2002) show differencesin processing product information depending on the manner of presentation of quan-tifier information. Hence descriptions such as “95% fat free” put the focus on the largeamount of non-fat in a product, whereas “5% fat” puts the emphasis on the fat content,even though these two expressions denote the same percentage of fat.

18. Hannon and Daneman (2004) used different materials, but this is in line with theirgeneral finding.

References

Adamson, Sylvia (1995). Empathetic narrative – A literary and linguistic problem. In Syntax and theLiterary System: New Approaches to the Interface between Literature and Linguistics, Wendy Ay-res-Bennett and Patrick O’Donovan (eds.), 17–42. Cambridge: Cambridge University Press.

Ali, Monica (2004). Brick Lane. London: Black Swan.Allen, Robert (2002). Punctuation. Oxford: Oxford University Press.Armstrong, Lance (2001). It’s Not About the Bike: My Journey Back to Life. London: Yellow Jersey Press.Atwood, Margaret (1992). The bog man. In Wilderness Tips, Margaret Atwood, 86–106. London: Virago.Baker, Linda and Jody L. Wagner (1987). Evaluating information for truthfulness: The effects of log-

ical subordination. Memory and Cognition 15: 247–255.Barthes, Roland (1974). S/Z, Richard Miller (trans.). New York: Hill and Wang.Barton, Stephen B. and Anthony J. Sanford (1993). A case study of anomaly detection: Shallow se-

mantic processing and cohesion establishment. Memory and Cognition 21: 477–487.Bray, Joe, Miriam Handley, and Anne C. Henry (eds.) (2000). Ma(r)king the Text: The Presentation

of Meaning on the Literary Page. Aldershot: Ashgate.


Breashears, David (2000). High Exposure: An Enduring Passion for Everest and Unforgiving Places.New York: Touchstone.

Bredart, Serge and Karin Modolo (1988). Moses strikes again: Focalization effect on a semantic illu-sion. Acta Psychologica 67: 135–144.

Broks, Paul (2004). Into the Silent Land. London: Atlantic Books.Brown, Kenneth A. (1994). Cycles of Rock and Water: Upheaval at the Pacific Edge. New York: Har-

perCollinsWest.Crystal, David and Derek Davy (1969). Investigating English Style. London: Longman.Dahl, Roald (1979 [1954]). Lamb to the slaughter. In Tales of the Unexpected, Roald Dahl, 22–31.

Harmondsworth: Penguin.Dean, Louise (2004). Becoming Strangers. London: Scribner.Dick, Philip K. (2001 [1969]). Ubik. London: Orion Publishing Group.Emmott, Catherine (1997). Narrative Comprehension: A Discourse Perspective. Oxford: Oxford Uni-

versity Press.— (2003). Reading for pleasure: A cognitive poetic analysis of “twists in the tale” and other plot

reversals in narrative texts. In Cognitive Poetics in Practice, Joanna Gavins and Gerard Steen(eds.), 145–159. London: Routledge.

— (2006). Reference: Stylistic aspects. In Encyclopedia of Language and Linguistics, 2nd ed., KeithBrown (ed.), vol. 10, 441–450. Oxford: Elsevier.

Emmott, Catherine, Anthony J. Sanford, and Lorna I. Morrow (2003). Towards a theory of readingin the age of cognitive science: Cross-disciplinary perspectives on narrative from stylistics andpsychology. Belgian Journal of English Language and Literatures New Series 1: 17–30.

— (2006). Sentence fragmentation: Stylistic aspects. In Encyclopedia of Language and Linguistics,2nd ed., Keith Brown (ed.), vol. 11, 241–251. Oxford: Elsevier.

Erickson, Thomas A. and Mark E. Matteson (1981). From words to meanings: A semantic illusion.Journal of Verbal Learning and Verbal Behavior 20: 540–552.

Fludernik, Monika (1993). The Fictions of Language and the Languages of Fiction. London/NewYork: Routledge.

Fox, Michael J. (2003). Lucky Man: A Memoir. London: Ebury Press.Franconeri, Steven L., Andrew Hollingworth, and Daniel J. Simons (2005). Do new objects capture

attention? Psychological Science 16(4): 275–281.Hannon, Brenda and Meredyth Daneman (2004). Shallow semantic processing of text: An individu-

al-differences account. Discourse Processes 37(3): 187–204.Heller, Zoë (2004). Notes on a Scandal. London: Penguin.Huddleston, Rodney and Geoffrey K. Pullum (2002). The Cambridge Grammar of the English Lan-

guage. Cambridge: Cambridge University Press.Ingarden, Roman (1973). The Literary Work of Art. Evanston: Northwestern University Press.Iser, Wolfgang (1978). The Act of Reading: A Theory of Aesthetic Response. Baltimore/London: The

John Hopkins University Press.King, Graham (2004). Good Punctuation. Glasgow: HarperCollins.Labov, William (1972). Language in the Inner City. Philadelphia: University of Pennsylvania Press.Lawrence, David H. (1996 [1928]). The woman who rode away. In The Woman who Rode Away and

other Stories, David H. Lawrence, 39–71. Harmondsworth: Penguin.Leech, Geoffrey and Michael H. Short (1981). Style in Fiction: A Linguistic Introduction to English

Fictional Prose. London/New York: Longman.Lessing, Doris (1966). A Proper Marriage. London: Grafton.Lively, Penelope (2004). The Photograph. London: Penguin.Longacre, Robert E. (1968). Discourse, Paragraph and Sentence Structure in Selected Philippine Lan-

guages. Volume I: Discourse and Paragraph Structure. Santa Ana, CA: The Summer Institute ofLinguistics.


McAteer, Erica (1989). Typeface effects in written language. Unpublished doctoral dissertation, Uni-versity of Glasgow.

— (1992). Typeface emphasis and information focus in written language. Applied Cognitive Psy-chology 6: 345–359.

Miller, Jim and Regina Weinert (1998). Spontaneous Spoken Language: Syntax and Discourse. Ox-ford: Clarendon Press.

Mills, Magnus (2003). The Scheme for Full Employment. London: Flamingo.Mukařovský, Jan (1964). Standard language and poetic language. In A Prague School Reader on Es-

thetics, Literary Structure, and Style, Paul L. Garvin (ed.), 17–30. Washington, D.C.: GeorgetownUniversity Press.

Niffenegger, Audrey (2005). The Time Traveler’s Wife. London: Vintage.Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (1985). A Comprehensive

Grammar of the English Language. London: Longman.Rabinowitz, Peter J. (1987). Before Reading: Narrative Conventions and the Politics of Interpretation.

Columbus: Ohio State University Press.Rankin, Ian (2000). The Hanging Garden. London: Orion.Sanford, Alison J. S., Anthony J. Sanford, Jo Molle, and Catherine Emmott (in press). Shallow

processing and attention capture in written and spoken discourse. Discourse Processes (Specialissue).

Sanford, Anthony J. (2002). Context, attention and depth of processing during interpretation. Mindand Language 17: 188–206.

Sanford, Anthony J. and Patrick Sturt (2002). Depth of processing in language comprehension: Notnoticing the evidence. Trends in Cognitive Sciences 6: 382–386.

Sanford, Anthony J., Nicholas Fay, Andrew J. Stewart, and Linda M. Moxey (2002). Perspective instatements of quantity, with implications for consumer psychology. Psychological Science 13(2):130–134.

Sebold, Alice (2002). Lucky. London: Picador.Short, Mick (2004). Graphological deviation, style variation and point of view in Marabou Stork

Nightmares by Irvine Welsh. In Stylistics: A Resource Book for Students, Paul Simpson (ed.),176–185. London: Routledge.

Shreve, Anita (1999). The Pilot’s Wife. London: Abacus.Simons, Daniel J. (2000a). Attentional capture and inattentional blindness. Trends in Cognitive Sci-

ences 4: 147–155.— (2000b). Current approaches to change blindness. Visual Cognition 7: 1–15.Simons, Daniel, J. and Daniel T. Levin (1997). Change blindness. Trends in Cognitive Sciences 1: 261–

267.Smith, Zadie (2003). The Autograph Man. London: Penguin.Sotirova, Violeta (2004). Conjunctions in free indirect style: Continuity or shift. Language and Lit-

erature 13(3): 216–234.Steen, Gerard (1994). Understanding Metaphor in Literature. London: Longman.Stubbs, Michael (2005). Conrad in the computer: Examples of quantitative stylistic method. Lan-

guage and Literature 14(1): 5–24.Sturt, Patrick, Anthony J. Sanford, Andrew Stewart, and Eugene Dawydiak (2004). Linguistic focus

and good-enough representations: An application of the change-detection paradigm. Psycho-nomic Review and Bulletin 11: 882–888.

Toolan, Michael (2004). Graded expectations: On the textual and structural shaping of readers’ ex-perience. In The Dynamics of Narrative Form: Studies in Anglo-American Narratology, JohnPier (ed.), 215–237. Berlin: Walter de Gruyter.

Townsend, Sue (1989). Rebuilding Coventry. London: Methuen.van Oostendorp, Herre and Sjaak de Mul (1990). Moses beats Adam: A semantic relatedness effect

on a semantic illusion. Acta Psychologica 74: 35–36.


van Peer, Willie (1986). Stylistics and Psychology: Investigations of Foregrounding. London: CroomHelm.

van Peer, Willie and Jèmeljan Hakemulder (2006). Foregrounding. In Encyclopedia of Language andLinguistics, 2nd ed., Keith Brown (ed.), vol. 4, 546–550. Oxford: Elsevier.

Wales, Katie (2001). A Dictionary of Stylistics. Harlow: Longman.Waugh, Nancy C. (1963). Immediate memory as a function of repetition. Journal of Verbal Learning

and Verbal Behaviour 2: 107–112.

Capturing the attention of readers? Stylistic and psychological ...

Documents

Transcript of Capturing the attention of readers? Stylistic and psychological ...